Explainable Artificial Intelligence (XAI) : a new trend for Autonomous Driving

Autonomous Driving is a field of application of Artificial Intelligence, and especially of Deep Learning.

But the fact is that Deep Learning cannot explain its behaviour. For Autonomous Driving, in case of accident, it is necessary to understand if the Autonomous Vehicle made a mistake and if yes, to know what this mistake is, why, and how to fix it.

Then it is necessary to develop a new kind of Artificial Intelligence system called : eXplanable Artificial Intelligence (XAI).


(figure by DARPA)

DARPA started last year a new research program on XAI (see : https://www.darpa.mil/program/explainable-artificial-intelligence ).

NEXYAD has been developing the only AI module in the world that computes driving risk in real time : twenty times per second, SafetyNex computes the risk that the "driver" is currently taking :
. If the driver is a human being, then it is possible to alert and let time to slow down before a potential accident
. If the driver is an Autonomous Driving Computer Program, then is is possible to use risk as a variable to servo control the Autonomous Driving system. This disrupts completely Autonomous Driving : no need to cross scenarios and situations ... risk is computed in real time and Autonomous Driving is ADAPTIVE to driving risk.

This module SafetyNex is a XAI.

It appears that XAI may be the next acronym that you will hear about in the next years.